[llvm-branch-commits] [llvm] [X86] Remove extra MOV after widening atomic load (PR #148898)

[Eli Friedman via llvm-branch-commits]

================
@@ -1204,6 +1204,13 @@ def : Pat<(i16 (atomic_load_nonext_16 addr:$src)), (MOV16rm addr:$src)>;
 def : Pat<(i32 (atomic_load_nonext_32 addr:$src)), (MOV32rm addr:$src)>;
 def : Pat<(i64 (atomic_load_nonext_64 addr:$src)), (MOV64rm addr:$src)>;
 
+def : Pat<(v4i32 (scalar_to_vector (i32 (zext (i16 (atomic_load_16 addr:$src)))))),
+           (MOVDI2PDIrm addr:$src)>;   // load atomic <2 x i8>
----------------
efriedma-quic wrote:

This looks wrong.  I think you need pinsrw or vpbroadcastw to load 16 bits.

https://github.com/llvm/llvm-project/pull/148898